American Journal of Chemical Engineering

2015; 3(2-1): 9-17

Published online January 16, 2015 (http://www.sciencepublishinggroup.com/j/ajche)
doi: 10.11648/j.ajche.s.2015030201.12
ISSN: 2330-8605 (Print); ISSN: 2330-8613 (Online)

Development of model equations for predicting gasoline

blending properties
M. K. Oduola1, 2, A. I. Iyaomolere2
1
Department of Chemical Engineering, University of Port Harcourt, Port Harcourt, Nigeria
2
Centre for Gas, Refining and Petrochemicals, Institute of Petroleum Studies, University of Port Harcourt, Port Harcourt, Nigeria

Email address:
koyejo.oduola@uniport.edu.ng (M. K. Oduola), adeiyaomo@cgrpng.org (A. I. Iyaomolere)

To cite this article:

M. K. Oduola, A. I. Iyaomolere. Development of Model Equations for Predicting Gasoline Blending Properties. American Journal of Chemical
Engineering. Special Issue: Developments in Petroleum Refining and Petrochemical Sector of the Oil and Gas Industry.
Vol. 3, No. 2-1, 2015, pp. 9-17. doi: 10.11648/j.ajche.s.2015030201.12

Abstract: Gasoline blending is of pertinent importance in refinery operations owing to the fact that gasoline gives about 60 -
70 % of the refinery profit. The blending process is essential to obtain gasoline in the demanded quantities and ensure property
specifications are met. Two model equations, multivariable nonlinear and multivariable exponential are proposed in this study
which are useful in predicting three significant properties of a motor gasoline: research octane number, reid vapour pressure and
specific gravity. Gasoline blend data obtained from four different streams: straight run gasoline, straight run naphtha, reformate
and fluidized catalytically cracked gasoline have been subjected to multivariate regression analysis with the aid of a statistical
software to ascertain the fitness of the proposed equations in predicting the research octane number, reid vapor pressure and the
specific gravity of the resulting premium motor spirit. The results of the regression analysis showed that the nonlinear
multivariable models proposed gave a good fit as evidenced by the value of the coefficient of determination R2 = 0.988 & 0.994
for the research octane number, 0.853 & 0.883 for the reid vapor pressure and 0.988 for specific gravity. In conclusion, the
proposed model equations were fit to the data, found to be adequate, and therefore could be used for prediction of the blend
gasoline properties.
Keywords: Gasoline Blends, Modeling, Petroleum Refining, Octane Number, Reid Vapor Pressure, Multivariate Regression

usually are sent to upgrading units (for octane improvement,

1. Introduction
Crude petroleum in its natural state has little or no use,
though it remains an essential source of energy affecting
nearly every aspect of our modern lifestyle; it has to be refined
in a refinery to different useful products. The refining process
aims to maximize the value added by separating the crude oil
using different physical and chemical processes into both
intermediate products which can serve as feedstock to other
downstream processes and finished products, including
transportation fuels [1]. Some of the products are Liquefied
Petroleum Gas (LPG), Straight Run Naphtha (SRN), Straight
Run Gasoline (SRG), Kerosene, Gasoline or Premium Motor
Spirit (PMS), Light Cycle Oil (LCO), Heavy Gas Oil (HGO),
Reformates, Alkylates, amongst others. Of these products,
PMS is the most important as it serves as fuel for most
automotive engines which has proliferated because of the
advancement in technology.
The naphthas are gasoline boiling range materials; they
sulfur control, etc.) and then to gasoline blending pool. The
distillates, including kerosene, usually undergo further
treatment and then are blended to jet fuel, diesel and home
heating oil. The gas oils go to conversion units, where they are
broken down into lighter (gasoline, distillate) streams. Finally,
the residual oil (or bottoms) is routed to other conversion units
or blended to heavy industrial fuel and/or asphalt. The bottoms
have relatively little economic value indeed lower value than
the crude oil from which they come. Most modern refineries
convert, or upgrade, the low-value heavy ends into more
valuable light products (gasoline, jet fuel, diesel fuel, etc.).
Conversion processes carry out chemical reactions that
fracture (crack) large, high-boiling hydrocarbon molecules
(of low economic value) into smaller, lighter molecules
suitable, after further processing, for blending to gasoline, jet
fuel, diesel fuel, petrochemical feedstocks and other
high-value light products. Conversion units form the essential
core of modern refining operations because they enable the
refinery to achieve high yields of transportation fuels and
10 M. K. Oduola and A. I. Iyaomolere: Development of Model Equations for Predicting Gasoline Blending Properties

other valuable light products, provide operating flexibility for different products in different volumes and with different
maintaining light product output in the face of normal qualities and these different products have to be mixed
fluctuations in crude oil quality, and permit the economic use optimally in the right volumes to obtain a product with desired
of heavy, sour crude oils [1-4]. properties in a blending process. Gasoline blending is the
Heavy gas oil from crude distillation is converted to light mechanical mixing of gasoline blending components streams
gases, petrochemical feedstock, gasoline blendstock (Fluid to obtain various grades of gasoline and its very important to
Catalytic Cracking (FCC) naphtha) and diesel fuel blendstock refinery economics as blending reduces property giveaway
(light cycle oil) in the FCC process which employs zeolite as [5-8]. PMS blending is a very important part of refinery
catalyst and operate at high temperature and low pressure. operations. Due to a large volume of the product, it is of
Distillates and heavy gas oils from crude distillation is pertinent importance to blend gasoline at the lowest possible
converted in the hydrocracking unit to light gases, cost, while satisfying quality constraints [9-11].
petrochemical feedstocks, gasoline and diesel fuel blendstocks. Seven most commonly employed refinery streams (cuts) in
The hydrocracking process is a catalytic process and operates the gasoline production are: fluidized catalytic cracking
at moderate temperature and high pressure. (FCC), reforming (REF), isomerization (ISO), alkylation
Upgrading processes carry out chemical reactions that (ALK) and dimersol (DIM), together with the butane's
combine or re-structure molecules in low-value streams to fraction (C4) and oxygenate additives such as MTBE or
produce higher-value streams, primarily high-octane, low ethanol. These components as well as the ranges of their
sulphur gasoline blendstock. The upgrading processes of volumetric content in the blends and of their RON values are
primary interest all employ catalysts, involve small presented in Table 1 [9,10].
hydrocarbon molecules, and apply to gasoline production.
Some important upgrading processes are catalytic reforming Table 1. Gasoline components employed for the blends preparation
in which catalytic reactions are carried out on straight run Min Max Min Max Typical
Fraction
naphtha from crude distillation to increase the octane number Vol % Vol % RON RON RON
of this stream and obtain premium, high-octane blendstock FCC 6.6 100.0 92.5 94.0 93.4
called reformate; alkylation which combines light olefins REF 0.0 54.4 98.2 103.6 101.4
(primarily C4 and some C3) with isobutane to produce a high ISO 0.0 32.0 77.5 86.2 84.5
octane gasoline blendstock called alkylate; isomerization
ALK 0.0 20.7 93.3 94.7 94.0
rearranges the low-octane C5 and C6 normal paraffin
molecules in light straight run naphtha to corresponding DIM 0.0 14.8 93.8 97.6 95.0
higher-octane C5 and C6 iso-paraffins thereby significantly C4 0.0 7.4 - - 92.0
increasing the octane of the resulting naphtha stream Ethanol 0.0 10.0 - - 116.0
(isomerate) and making it a valuable gasoline blendstock and
Etherification combines C4 and/or C5 olefins produced by Gasoline revenues accounts for a good share of the
FCC plants with a purchased alcohol (methanol or ethanol) to profitability of refineries. Mendez et al. [12] reported that
produce an ether (a class of oxygen-containing organic gasoline production often yields 60-70% of a typical refinerys
compounds). Ethers are premium gasoline blendstocks, with total revenue. The octane number of gasoline is an indication
very high octane and other desirable blending properties [1]. of its quality and a higher octane rating gasoline is more
Gasoline which is a complex mixture of hydrocarbons desirable and more expensive than a lower octane rating
contains paraffins, olefins, napthenes and aromatics was gasoline. It is therefore economically desirable to blend to get
originally discarded as a by-product of kerosene production a desired optimum gasoline with favorable good price and also
but its ability to vaporize at low temperatures makes it useful to improve the quality of the gasoline.
as a fuel for many machines. Gasoline contains hydrocarbons PMS blending could be tedious and time consuming. Lots
having boiling points in the range between 30oC to 215oC. of experiments are needed to determine the individual
Gasoline was first produced initially by atmospheric properties of the blend components and that of the final PMS
distillation but the product obtained from this process had low and these components could be as over 100 in some cases
qualities like the Research Octane Number (RON); which is [13,14]. Given a set of existing data, the blend behavior of the
an indication of the knock resistance of the gasoline in engines, gasoline components can be predicted by representing the
and the volume obtained could not meet up with increasing process via a mathematical model [15]. Mathematical model
demand. This dearth led to increased research into production for gasoline blending is the mathematical presentation of
of gasoline from other less commercial refinery products in gasoline blend components to allow investigation of key
several chemical processes and an attendant improvement in properties of the system and prediction of future outcomes to
refinery complexity. be carried out [16]. Without accurate blending correlations,
The ever increasing demand for gasoline makes it any attempt to blend different gasoline cuts can be expected to
imperative to ensure that this product is readily available in the achieve non-optimum results [17]. Hence, there is need to
demanded volume while ensuring that its quality and obtain a mathematical model to effectively predict these
specifications needed for effective performance in car engines properties of the PMS from various blendstocks in order to
is not compromised. The different refinery processes produce improve refinery profitability
 American Journal of Chemical Engineering 2015; 3(2-1): 9-17
reduce property giveaway
reduce the cost of performing several experiments to
determine PMS components and product properties
thereby saving costs
reduce the time needed to blend a batch and ensure
continual availability of gasoline [18].
Research and Motor Octane numbers (RON, MON), Reid
Vapor Pressure (RVP) and Specific Gravity (SG) are the main
quality characteristic of the gasoline, as they provide a
sensitive indication of the anti-knocking behavior of the fuel.
The higher the octane number the better the gasoline resists
detonation and the smoother the engine runs. Other important
technological properties of the commercial gasoline are
ASTM distillation points, and sulfur content. These
properties are monitored during production to ensure the
required technological and environmental quality level of the
final gasoline is gained. Although the final gasoline has to
meet all the product specification, RON and MON are
considered to be the most important. The Reid Vapour
Pressure is an indication of the volatility of gasoline in internal
combustion engines while the Specific Gravity gives an
indication of how heavy the product is. RON and MON values,
as well as most of the gasoline properties, blend in non-ideal
fashion, i.e. they depend non-linearly on the mixture
composition. Therefore algorithms more complex than linear
combination are necessary for the reliable prediction. There is
a marked interest in refineries to utilize algorithms for the
prediction of the octane rating of the gasoline blends [19].
Numerous studies in the past have attempted to
mathematically describe the octane number as a function of
the gasoline composition. Schoen and Mrstik [20] developed a
graphical correlation for predicting octane numbers of blends
as a series of binary systems based on the octane rating and
volumetric olefin contents of the two components being
blended. The blended octane number yields different values
depending upon the order of calculation. Stewart [21] refined
this method to be applicable to multicomponent blends
yielding more self-consistent results. Stewarts correlation
also required the octane rating and volume percent olefins of
the components being blended. Auckland and Charnock [4]
developed a blending index to blend octane number linearly.
The blending index is a hypothetical value obtained by
extrapolating from the octane rating at a given concentration
to an octane rating at 100% concentration of component.
However, obtaining blend values by blending the blending
indices linearly is analogous to obtaining the molar property
of a real solution by a linear combination of the partial molar
properties of its components. Since the partial molar property
depends on the composition, so does the blending index.
Therefore, the usage of this approach is limited. The index
method can only be used to find the blending value of a
component at a particular composition and cannot be used to
predict its blending values in other mixtures.
Rusin et al. [22] presented a rather complicated
transformation method for estimating the octane number of
gasoline blends from the octane ratings of the components,
their concentrations, and contents of olefins, aromatics and
11
paraffins. The method consists of three steps: (a)
transformation of component properties (b) linear blending of
these transformed properties, and (c) inverse transformation of
the results. This method is similar to the blending index
method. Due to the back and forth transformation, this method
may also cause inconsistency in data transformation between
these three steps. Healy et al [10] correlated gasoline
component blending with differences in octane level and
hydrocarbon type among components. However, if the
equation of Healy et al. is used to predict the blending
behavior of new components, unreasonable blending values
may be obtained, especially if the hydrocarbon type or octane
number of the new component is outside the range of the
component previously tested [23].
An interesting equation was proposed by Morris et al. [23]
for describing nonlinear gasoline blending behavior as
follows:
= + + (1)
where ai and xi are the octane number and the volume fraction
of component i, respectively and b12 is the interaction
coefficient for components 1 and 2.
Equation (1), which represents a simple excess octane
number function, is quite effective in correlating the octane
numbers of gasoline blends. However, there are some
disadvantages of using this model to describe gasoline
blending behavior. First, the parameter b12 is an empirical
constant, which depends on the characteristics of components
1 and 2. The interaction parameter b12 not only depends on the
component types, but it also depends on the octane levels and
octane differences. The values of b12 can vary from large
negative to large positive ones and the interaction parameter
b12 must be changed if the component characteristics change.
The set of binary parameters obtained from equation (1) can
only be used for the specific set of components where the
parameters were derived. Therefore it would be very difficult
to generalize these interaction parameters as a function of
these properties for the purpose of prediction. A second
disadvantage is that when new components are added to the
gasoline pool, new interaction parameters are required to
describe the behavior of the new component with the present
components. Because it is non-trivial to run a blending study,
it is very costly to run additional blending studies whenever
there are new components to be blended into the gasoline
pool.
Pasadakis et al. [24] used Artificial Neural Network (ANN)
models have to determine the RON of gasoline blends
produced in a Greek refinery. The developed ANN models use
as input variables the volumetric content of seven most
commonly used fractions in the gasoline production and their
respective RON values. The predicting ability of the models,
in the multi-dimensional space determined by the input
variables, was thoroughly examined in order to assess their
robustness. Based on the developed ANN models, the effect of
each gasoline constituent on the formation of the blend RON
value was revealed.
Zahed et al. [25] proposed a model with five independent
12 M. K. Oduola and A. I. Iyaomolere: Development of Model Equations for Predicting Gasoline Blending Properties

! "# $ %!%&
variables for predicting the octane number of gasoline blends. = (3)
' &( ) ! " *& +!
The variables of the model can be regressed so that the model
works well for the blends used in the regression but it will still For the modeling equation, the volume fraction of each cut
perform poorly for other blends. The five variables derived (FCCG, SRN, SRG and REF) in the resulting PMS blend were
from this model are internally inconsistent. The octane taken as the dependent variables and represented by , ,
number of n-heptane predicted from this model is 108.77 , and respectively, while the dependent variable ,
versus the defined value of zero. Similarly, the octane number represents the resulting particular PMS properties being
of iso-octane predicted from this model is -108.95 versus the modeled (RON, RVP or SG). A DataFit regression package
defined value of 100. The set of variables obtained from this from Oakdale Engineering Software was used to fit the data
model can only be used for this specific set of components obtained [27].
within the same range of compositions where the variables
were derived. The use of this model to predict the octane
number of blends at other conditions will therefore be very
unreliable.
Prasenjeet et al. [26] presented a model that predicts the
octane numbers of a wide variety of gasoline process streams
and their blends including oxygenates based on detailed
composition. The octane number is correlated to a total of 57
hydrocarbon lumps measured by gas chromatography. The
model is applicable to any gasoline fuel regardless of the
refining process it originates from. It is based on the analysis
of 1471 gasoline fuels from different naphtha process streams
such as reformates, cat-naphthas, alkylates, isomerates,
straight runs, and various hydroprocessed naphthas. The
model predicts the octane number within a standard error of 1
number for both the research and motor octane numbers.
This study has been carried out with the objective to
develop a mathematical model for blending gasoline and
employ the model to predict predicting the research octane
number, reid vapour pressure and the specific gravity of
gasoline blends. This work attempts to develop a model for
predicting the blend properties of various gasoline cuts
Figure 1. Diagrammatic representation of the gasoline blender
obtained from the Port Harcourt Refinery Company Limited,
Nigeria. Table 2. PMS Blend data

Content Tank S/ Volume, SG RVP RON

2. Methodology No L
Fluid 52TK03 1 7800 0.7604 0.46 95.00
Figure 1 represents a typical industrial blending process Catalytic 2 7700 0.7601 0.56 95.10
employed to produce a quality gasoline from various product Cracking 3 7700 0.7584 0.61 94.7
streams. Five PMS blending data (Table 2) for gasoline Gasoline 4 8000 0.7589 0.47 94.8
(FCCG) 5 7500 0.7594 0.48 94.70
produced from four different blend cuts: Straight Run
Gasoline (SRN), Straight Run Naphtha (SRN), Reformate Straight 51TK01 1 2200 0.7697 0.15 60.00
Run 2 3100 0.7703 0.15 61.00
(REF) and Fluidized Catalytically Cracked Gasoline (FCCG) Naphta 3 3500 0.7708 0.14 61.0
were obtained by the Production Programming and Quality (SRN) 4 3000 0.7708 0.14 61.0
Control (PPQC) department of the Port Harcourt Refining 5 3800 0.7682 0.06 64.60
Company (PHRC), Nigeria. The raw gasoline cut qualities Straight 52TK01 1 2300 0.6519 1.17 73.50
(RON, RVP and SG) were determined by the Quality Control Run 2 1200 0.6599 1.05 76.00
department prior to blending using standard ASTM analytical Gasoline 3 500 0.6585 1.11 75.0
methods. (SRG) 4 1000 0.6549 1.11 74.0
The Production Programming Unit of the PPQC in PHRC 5 1200 0.6658 0.99 88.50
utilizes linear volumetric blending formula in formulating Reformate 52TK02 1 7700 0.7534 0.29 95.90
(REF) 2 8000 0.7512 0.28 96.30
gasoline blendstocks. The formula is presented below
3 8300 0.7568 0.40 97.0
= (2) 4 8000 0.7579 0.36 95.6
5 7500 0.7560 0.32 96.00
where represents the final RON of the PMS TOTAL VOLUME 20000
represents the volumetric ratio of the blend components
represents the RON of the blend components
 American Journal of Chemical Engineering 2015; 3(2-1): 9-17 13

3. Results and Discussion data employed for regression analysis based on the RON
blend data.
In this study, we employ the regression model technique for
the situation where we have measured five responses y1, y2, ... Table 3. Regression analysis data for RON blend data
y5 and the same set of 4 predictors x1 ... x4 on each sample unit.
Y x1 x2 x3 x4
Each response follows its own regression model, as an
example, for the nonlinear multivariable model we have the 89.0965 0.375 0.19 0.06 0.375
following equations: 89.01 0.4 0.15 0.05 0.4
89.2645 0.385 0.175 0.025 0.415
y1 = a01 + a11x1 + ... + a41x4 + 1
89.1485 0.385 0.155 0.06 0.4
y2 = a02 + a12x1 + ... + a42x4 + 2 89.024 0.39 0.11 0.115 0.385

...
Amongst several mathematical models worked upon,
... particularly noteworthy and hereby reported are the following
model equations:
y5 = a05 + a15x1 + ... + a45x4 + 5 (4) Model 1
where = (1, ...., 5) are the errors associated with different ,= + + +- . (5)
responses on the same sample unit, which have different
variances and may be correlated. and Model 2
Knowing from the basis of definition that x4 could be (6)
,= / + 0( ) + 0( ) + . 0( . )
expressed as a function of other independent variables, (4)
could be simplified further to yield (5). The dependent The key results of the regression analysis obtained with the
variables were obtained from the given blend data (Table 2) aid of the Datafit Regression Package for all the blend
using (3). The five sets of dependent variables corresponding properties considered in this study are summarized in Table 4.
to the values of the blend properties given in the data. Similar The regression coefficients corresponding to each model
procedure have been observed for the three parameters being equation are shown together with the values of the coefficient
modeled, the RON, RVP and SG. Table 3 shows the typical of determination (R2), and the standard error of estimate ().
Table 4. Regression analysis values for Models 1 and 2

Regression coefficients*
Blend property Model equation R2 Standard error,
3 , 35 6, 37 8, 39 :, 3;
RON Model 1 80.33 90.33 90.47 97.01 0.994 0.016
Model 2 119.2 -11.5 -5.79 -6.06 0.987 0.024
RVP Model 1 -1.35 -0.44 0.50 2.47 0.883 0.022
Model 2 9.03 -2.59 -2.48 -1.79 0.853 0.025
SG Model 1 0.82 0.77 0.62 0.70 0.988 0.0009
Model 2 0.64 0.084 0.06 -0.07 0.988 0.0009

* a, b, c, d correspond to Model 1, while /, , , . Model 2

As can be seen from Table 4, the coefficient of MON.

determination R2 for the RON data using Model 1 has a value
of 0.994 with a standard error of estimate of 0.016. This
implies that the data obtained from PHRC closely fitted to that
obtained from the model generated and that 99.4 percent of the
data set is accounted for by Model 1. Similar results obtained
with the aid of Model 2 (R2 = 0.988, = 0.024) confirmed the
applicability of this exponential model in describing the given
set of data. It follows that both the nonlinear multivariate
model equation and the exponential equation are very suitable
for predicting the behavior of RON of gasoline blends.
Increasing research into the motor and research octane number
of gasoline has been caused ever since the use of Lead as an
additive to motor fuel in the form of tetraethyl lead had been
progressively phased-out worldwide due to emission problem.
The anti knock index (AKI) is an average value of RON and
A graphical representation comparing the industry data
from PHRC and that predicted from Model 1 is shown in Fig.
2. Here, data and model plots are made on one graph; the trend
line of data obtained from PHRC is shown in dotted lines
while that predicted by the model is shown in thickened line. A
visual inspection shows that there is a close correlation
between the raw data and the model predicted data.
Similar trends were observed upon analysis of the
regression data for the other parameters, reid vapor pressure
and specific gravity, as can be seen from Table 4 and Figs 3
and 4. Sufficiently high R2 values (0.883 and 0.853 for RVP
using Models 1 and 2 respectively) with still higher values of
this statistical parameter (0.988 and 0.988 for SG models 1
and 2) further buttressed the fact that the proposed models can
adequately be employed for the target purpose.
14 M. K. Oduola and A. I. Iyaomolere: Development of Model Equations for Predicting Gasoline Blending Properties

Figure 2. Plot of RON fit curve for model 1 upward and model 2 downward.

Figure 3. Plot of RVP fit curve for model 1 left and model 2 right.

Figure 4. Plot of SG fit curve for model 1 upward and model 2 downward.
 American Journal of Chemical Engineering 2015; 3(2-1): 9-17 15

components like butane, alkylate, dimate, as well as "octane

4. Conclusion
Model equations have been put forward to predict final
properties, in particular the research octane number, reid vapor
pressure and specific gravity of the gasoline resulting from
blending of various gasoline cuts. The results of the regression
analysis showed that the nonlinear multivariable models
proposed gave a good fit as evidenced by the value of the
coefficient of determination R2 = 0.988 & 0.994 for the
research octane number RON, 0.853 & 0.883 for the reid
vapor pressure RVP and 0.988 for SG. The scope of the study
could be expanded to include other refinery gasoline blend
booster" gasoline additives including ethyl tert-butyl ether
(ETBE) and methyl tert-butyl ether (MTBE), isooctane,
toluene, and oxygenates where they are being utilized.
Acknowledgements
The contribution of the Management & Staff of the Port
Harcourt Refinery Company (PHRC), Port Harcourt,
NIGERIA to the success of this paper is gratefully
acknowledged, especially for the raw data on gasoline blends.

Appendix
RON Fit Information for Model 1

Model Definition: y = ax1+bx2+cx3+dx4

Number of observations = 5 Average Residual = -8.5265128291212E-15

Number of missing observations = 0 Residual Sum of Squares (Absolute) = 2.56784042317984E-04

Solver type: Nonlinear Residual Sum of Squares (Relative) = 2.56784042317984E-04

Nonlinear iteration limit = 250 Standard Error of the Estimate = 1.60244825912721E-02

Diverging nonlinear iteration limit =10 Coefficient of Multiple Determination (R^2) = 0.9940174678

Number of nonlinear iterations performed = 11 Proportion of Variance Explained = 99.40174678%

Residual tolerance = 0.0000000001 Adjusted coefficient of multiple determination (Ra^2) = 0.9760698712

Sum of Residuals = -4.2632564145606E-14 Durbin-Watson statistic = 3.18736483829897

Regression Variable Results

Variable Value Standard Error t-ratio Prob(t)

A 80.3334859739025 0.932619716219116 86.13745193 0.00739

B 90.3313258441462 0.354588591414152 254.7496677 0.0025

C 90.4714651835739 0.463398896931756 195.2345286 0.00326

D 97.0094259621047 0.825670550622306 117.4916871 0.00542

99% Confidence Intervals

Variable Value 99% (+/-) Lower Limit Upper Limit

A 80.3334859739025 59.3682395852184 20.9652463886841 139.701725559121

B 90.3313258441462 22.5722232579464 67.7591025861998 112.903549102093

C 90.4714651835739 29.4988152814333 60.9726499021406 119.970280465007

D 97.0094259621047 52.5601230762394 44.4493028858653 149.569549038344

Variance Analysis

Source DF Sum of Squares Mean Square F Ratio Prob(F)

Regression 3 0.04266551595768 0.01422183865256 55.38443325 0.09838

Error 1 2.56784042317984E-04 2.56784042317984E-04

Total 4 0.042922299999998
16 M. K. Oduola and A. I. Iyaomolere: Development of Model Equations for Predicting Gasoline Blending Properties

RVP Fit Information Model 1

Equation ID: Model 1

Model Definition: y = ax1+bx2+cx3+dx4
Number of observations = 5 Average Residual = -1.88737914186277E-16
Number of missing observations = 0 Residual Sum of Squares (Absolute) = 4.84349755480196E-04
Solver type: Nonlinear Residual Sum of Squares (Relative) = 4.84349755480196E-04
Nonlinear iteration limit = 250 Standard Error of the Estimate = 2.20079475526501E-02
Diverging nonlinear iteration limit =10 Coefficient of Multiple Determination (R^2) = 0.883210643
Number of nonlinear iterations performed = 11 Proportion of Variance Explained = 88.3210643%
Residual tolerance = 0.0000000001 Adjusted coefficient of multiple determination (Ra^2) = 0.5328425722
Sum of Residuals = -9.43689570931383E-16 Durbin-Watson statistic = 3.18736483829746
Regression Variable Results
Variable Value Standard Error t-ratio Prob(t)
A -1.34885700506242 1.28085544629671 -1.053090736 0.48354
B -0.437995410645829 0.486990271178 -0.899392527 0.53369
C 0.503133312444704 0.636429823024969 0.790555839 0.57413
D 2.47296374594515 1.1339719750927 2.180797939 0.27371
99% Confidence Intervals
Variable Value 99% (+/-) Lower Limit Upper Limit
A -1.34885700506242 81.536055572633 -82.8849125776954 80.1871985675705
B -0.437995410645829 31.0005831875135 -31.4385785981594 30.5625877768677
C 0.503133312444704 40.5135314592119 -40.0103981467672 41.0166647716566
D 2.47296374594515 72.1858210044637 -69.7128572585185 74.6587847504088
Variance Analysis
Source DF Sum of Squares Mean Square F Ratio Prob(F)
Regression 3 3.6628582445198E-03 1.22095274817327E-03 2.52080802 0.4265
Error 1 4.84349755480196E-04 4.84349755480196E-04
Total 4 0.004147208

SG Fit Information for Model 2

Equation ID: Model 2

Model Definition: y = a0+a1exp(x1)+a2exp(x2)+a3exp(x3)
Number of observations = 5 Average Residual = 7.99360577730113E-16
Number of missing observations = 0 Residual Sum of Squares (Absolute) = -1.88737914186277E-16
Solver type: Nonlinear Residual Sum of Squares (Relative) = 8.77507022601528E-07
Nonlinear iteration limit = 250 Standard Error of the Estimate = 9.36753448139652E-04
Diverging nonlinear iteration limit =10 Coefficient of Multiple Determination (R^2) = 0.9876247106
Number of nonlinear iterations performed = 10 Proportion of Variance Explained = 98.76247106%
Residual tolerance = 0.0000000001 Adjusted (Ra^2) = 0.9504988423
Sum of Residuals = 3.99680288865056E-15 Durbin-Watson statistic = 3.24519825698429
Regression Variable Results
Variable Value Standard Error t-ratio Prob(t)
a0 0.636780770242741 0.180899312307499 3.520083974 0.17621
a1 8.39111049388251E-02 7.02630414986109E-02 1.194242423 0.44379
a2 6.00086191331493E-02 0.042902232857076 1.39872951 0.39514
a3 -7.28034857759079E-02 3.40478454804993E-02 -2.138269977 0.27849
99% Confidence Intervals
Variable Value 99% (+/-) Lower Limit Upper Limit
a0 0.636780770242741 11.5155979732146 -10.8788172029719 12.1523787434574
a1 8.39111049388251E-02 4.47276956419783 -4.388858459259 4.55668066913665
a2 6.00086191331493E-02 2.73104888809931 -2.67104026896616 2.79105750723246
a3 -7.28034857759079E-02 2.16740072367489 -2.24020420945079 2.09459723789898
Variance Analysis
Source DF Sum of Squares Mean Square F Ratio Prob(F)
Regression 3 7.00304929773986E-05 2.33434976591329E-05 26.60206364 0.14135
Error 1 8.77507022601528E-07 8.77507022601528E-07
Total 4 7.09080000000001E-05
 American Journal of Chemical Engineering 2015; 3(2-1): 9-17
References
[1] MathPro, An introduction to petroleum refining and the
production of ultra low sulfur gasoline and diesel fuel, the
International Council of Clean Transportation (ICCT), 2011.
[2] C.Y. Jaja, Recent trends and patterns of gasoline consumption
in Nigeria, Council for the Development of Social Science
Research in Africa, pp. 159 177, 2010.
[3] J. Wodausch, Investigation and prediction of autoignition
during hot start conditions, M.Sc. Thesis, Nelson Mandela
Metropolitan University, 2009.
[4] M.H.T. Auckland and D.J. Charnock, The Development of
Linear Blending Indices for Petroleum Properties, J. Inst. of
Petroleum, 55(545), 322- 329, 1969.
[5] M.B. Celik, Experimental determination of suitable
ethanol-gasoline blend rate at high compression ratio for
gasoline engine, Appl. Thermal Engrng., .28, 396404, 2008.
[6] A.E. Eiman, Improvement of gasoline octane number by
blending gasoline with selective components, M.Sc Thesis,
University of Technology, Republic of Iraq, 2008.
[7] M.M. Farsibaf, M. Golchinpour, and A. Barzegar, Global
optimization in order to find blend composition of gasoline of
desired octane number considering ethanol as octane-booster,
41st International Conference on Computers & Industrial
Engineering, 313 - 318, 2012.
[8] S. Fernando, Internal combustion engines, PhD Thesis,
Department of Aerospace and Mechanical Engineering
University of Notre Dame, Notre Dame, 1998.
[9] A. Hassan and Mourhaf, Correlation between the octane
number of motor gasoline and its boiling range, J. King Saud
Univ., Vol. 9, Eng .Sci. (2), pp. 311-318, 1996.
[10] Jr. W.C. Healy, C.W. Maassen, and R.T. Peterson, A new
approach to blending octanes, API Division of Refining, 24th
midyear meeting, New York, 1959.
[11] W.G. Lovell, J.M. Campbell, T.A. Boyd, Detonation
characteristics of some aliphatic olefin hydrocarbons, Ind. Eng.
Chem.. 555, 1931.
[12] C.A. Mendez, I.E. Grossmann and I. Harjunkoski,
Optimization techniques for blending and scheduling of
oil-refinery operations, Carnegie Mellon University Pittsburgh,
U.S.A. 2011.
[13] D.S. Monder, Real-time optimization of gasoline blending with
uncertain parameters, M.Sc Thesis, University of Alberta,
Canada, 2001.
17
[14] G. Najafi, B. Ghodadiam, T. Tavakoli, D.R. Buttsworth, T.F.
Yusaf and M. Faizollahneja, Performance and exhaust
emissions of a gasoline engine with ethanol blended gasoline
fuels using artificial neural network, Ind. Eng. Chem. Res,
86(5), pp. 630639, 2009.
[15] I.L. Nwaogazie, Probability and statistics for science and
engineering practice, 3rd Edition, De-Adroit Innovation,
Nigeria, 2011.
[16] O.A. Francis, A model for blending motor gasoline, case study:
Tema Oil Refinery Ltd., M.Sc. Thesis, Kwame Nkrumah
University of Science and Technology, 2013.
[17] Chorng H. Twu and John E. Coon, A Generalized Interaction
Method for the Prediction of Octane Numbers for Gasoline
Blends. Simulation Sciences Inc., 601 South Valencia Avenue,
Brea, CA 92621 (USA).
[18] http://www.osha.gov, OSHA Technical Manual. Accessed on
18th October, 2014 .
[19] E. Paranghooshi and M. T. Sadeghi, Predicting octane numbers
for gasoline blends using artificial neural networks,
Hydrocarbon Processing, 2009.
[20] W.F. Schoen and A.V. Mrstik, Calculating gasoline blend
octane ratings, Ind. and Engr. Chem., 47(9), 1740-1742, 1955.
[21] W.E. Stewart, Predict Octanes for Gasoline Blend, Petroleum
Refiner, 38(12), 135-139, 1959.
[22] M.H. Rusin, H.S. Chung, and J.F. Marshall, A transformation
method for calculating the research and motor octane numbers
of gasoline blends, Ind. Eng. Chem. Fundam., 20(3), pp.
195-204, 1981.
[23] W.E Morris, Interaction Approach to Gasoline Blending,
NPRA Paper AM-75-30, National Petroleum Refiners
Association annual meeting, 1975.
[24] N. Pasadakis, V. Gaganis, Ch. Foteinopoulos: Octane number
prediction for gasoline blends, Fuel Processing Technology,
87, (2006) 505-509.
[25] A. H. Zahed, S. A. Mullah, and M. D. Bashir. Predict Octane
Number for Gasoline Blends. Hydroc Proc, 72(5), 1993, 85-87.
[26] G. Prasenjeet, J.H. Karlton, and B.J. Stephen, Development of
a detailed gasoline composition-based octane model, Ind. Eng.
Chem. Res., pp. 337-345, 2006.
[27] http://www.curvefitting.com, DataFitX version 8.2.79.